v0.8.1

v0.8.1 is a re-run of the definitive v0.8.0 run with the detection changes below. Of the 23,570 genomes with a report in both runs, 23,214 are identical in calls, confidence, coordinates and gene models and 356 differ; every difference was traced to a listed change. No call is lost and no idiomorph flips. Full details: CHANGELOG.

Results of the definitive run

Code and database as of commit 50bc950 (identical src/, db/ and scripts/ to this tag), --preset v0.8.1, 110 jobs, all completed.

Campaign	Genomes	Called	Uncalled	No report (suppressed)	Loci (high / medium / low)
Ascomycota	19,415	17,618	1,725	72	20,843 (14,623 / 6,120 / 100)
Basidiomycota	3,275	2,995	268	12	5,993 (3,822 / 2,161 / 10)
Mucoromycotina	974	868	105	1	914 (identical to v0.8.0)

A determinism check (60 genomes run twice) gave no differences in family, idiomorph, confidence, contig or coordinates.

What changed in detection

• Fallback after withheld hits (#228). New option --fallback-after-withheld-policy on (off by default; on in the new preset v0.8.1): when the routed family's hits were all withheld, the phylum-fallback families (MATsc, MTL) are searched too. 17 genomes are newly called (7 Ascomycota, 10 Basidiomycota).
• S. pombe silent cassettes (#254, #274). A merged mat1 call holding a P and an M cluster 5-20 kb apart is split into a mat2 P and a mat3 M call (37 genomes). The #252 tie-break is reverted.
• Rust allow list (#165, #269). db/allow.txt lets nine rust genomes (Austropuccinia psidii, Hemileia vastatrix, Phakopsora pachyrhizi; 0.56-1.27 Gb) be searched: eight are called (rustHD, 14 loci).
• Hyphopichia MTLA1 record (#148, #273). A candidate CTG-clade MTLA1 record changes 190 Serinales genomes in a call field (4 new calls, 160 span changes, 36 loci up and 14 down in tier); the drops are the 39-bit noise floor tracked in #149.
• Report-only additions (no call changes). resolution_flags (contig_edge, partial_core) on every locus (#257); two_idiomorphs for Basidiomycota MAT and redPR (#271); a cross-family cassette indication with shared-flank measurements (#264).

Site and tools

• The website at https://mat.fungalgenomes.org/ shows the v0.8.1 data and counts (22,979 NCBI genomes, 20,871 called, 27,103 loci).
• Browser tool: failed downloads are retried; the database loads as one archive; .fsa, .seq and .dna uploads; a Reset and a Copy log button; a "Guess phylum with GmG" button.
• matpredict detect --phylum auto asks GmG which phylum a genome is in.
• Raw-reads typing: contamination screen, UNVALIDATED species gate, sortable species tables.

Known limits (tracked, not changed in this release)

• #275: with an unpolished short MAT1-2-1 hit and a polished MAT1-1-3 model overlapping, the first-pass identity can pick the wrong idiomorph (seen in an assembly-rescue run of Trichoderma viride).
• #276: tblastn -seg no makes a few genomes more than an hour slow; a faster, call-preserving setting is being tested for v0.9.0.
• #149, #145, #138, #129: confidence and idiomorph-vote questions for weak fragments (v0.9.0).

Install

• Docker image ghcr.io/stajichlab/matpredict:0.8.1 (also 0.8 and latest), built when this release is published.
• Source: this tag. Cite with CITATION.cff (version 0.8.1, released 2026-11-06).